6

Several early computers (e.g., KDF9, LGP30) used Friden Flexowriters for console functions.

What was the nature of the connection between the Flexowriter and the computer? Was this something we would recognize as a "serial interface" today?

As well as the usual keyboard and printer functions, Flexowriters typically had integrated papertape readers and punches. Were these controlled by inband signaling (like XON/XOFF for ASCII Teletypes; not that I'm saying a Flex used ASCII) or "extra wires"?

Flexowriters also seemed to have switches, some of which were clearly for local device control (start tape, perhaps) but others seem to be computer control (the LGP30 Flex had what looks like a "start computer" switch). How did they communicate with the computer?

If any answer is computer-specific, please indicate the computer you're talking about.

4 Answers 4

4

Partial answer:

While I wasn't able to find schematics for how the Flexwrite is wired up to the KDF9 (say), there are some Flexwriter manuals availavle on bitsavers. The Flexwriter apparent came with 5 bit and 8 bit encodings (or "5-channel" and "8-channel" variants, as they call them).

The encodings (at least the earlier ones used for the KDF9) were Friden-specific, and neither ASCII nor compatible to anything else. However, the 5-bit encoding worked similar to other Teletext-like encodings (there are shift-states for upper and lower case, and the "all holes punched" encoding is a null encoding).

The Communications System and Equipment (1961) brochure suggest that serial communication was in use, also over longer distances. (Though that doesn't exclude the possibility the Flexwriters connected directly to the computer used a parallel connection).

There is no indication of any software flow control (XON/XOFF). There is also no indication of character codes which could be used to change channels etc. as part of the transmitted data, so I'd assume this could only be done on the flexwriter.

In various materials it is mentioned that the Flexwriters operated at a fixed speed.

There is also plenty of material on the KDF9. This here describes the I/O system as using DMA, that's probably why software flow control wasn't necessary.

It also mentions that there was a dedicated Flexwriter used as the console for the KDF9 which had an "attention" key that was hardwired as an interrupt source, and this feature seems to have been unique for this special Flexwriter. So that was probably a custom modifications.

All I/O channels of the KDF9 work similarly, so likely other Flexwriters were just hooked up to a single channels like the console.

Edit

Looking at the PDP-1 Maintenance Manual, it says on page 4-4

(c) The typewriter logic cable is equipped with a 50-pin Cannon connector

and on page 9-10 ff.

The PDP-l typewriter is an IBM Model B equipped with a Soroban electromechanical encoder and decoder. The decoder contains six information solenoids. These solenoids are driven from the typewriter control unit. The positions of the solenoid armatures are mechanically decoded to determine the desired character. Besides providing the six information signals, typewriter control also energizes the typewriter cam magnet, thereby causing the typewriter to print the desired character. The typewriter decodes most control characters (carriage return, backspace; space, tab, and shift), and all print characters, from the six information solenoids. The two color characters, red and black, are decoded by the typewriter control unit, which directly controls the color shift solenoid in the typewriter.

When the operator strikes a typewriter key, the typewriter encoder presents information signals to the typewriter control unit. The encoder mechanically encodes all print characters into six coded switch closures plus a common switch closure. Control characters are not encoded at the typewriter. When a control character key is struck, a single switch is closed. Closure of such a control char'acter switch is encoded within typewriter control.

So this looks very much like a parallel connection, very likely 5 bits plus parity.

7
  • 1
    Thanks. I had poked around bitsavers a little, but had not found the communications brochure. I'll take a look at it. Having used KDF9, including prep of Algol 60 code on an (offline) Flex, I'm familiar with the coding issues, FWIW, my mention of XON/XOFF was not about flow control, but of papertape reader operation, for comparison with the ASR33 Teletype. Jun 8, 2019 at 15:44
  • 1
    I don't know of any KDF9 that had "other" Flexowriters apart from the console. Did any such exist? The multi-access systems I'm aware of tended to use teletypes connected to a front-end multiplexer (a PDP8 in at least 2 cases). I myself am a former user of Eldon2 at Leeds. But here I'm straying from my own topic. Jun 8, 2019 at 15:49
  • 1
    Aha, excellent info from the PDP-1 Maintenance Manual. Your bitsavers-fu is clearly better than mine. Jun 8, 2019 at 15:52
  • 1
    I didn't know about the front-end multiplexer, I'd just assumed that additional teletypes for multi-access system would be connected to other channels. But I'm sure there was a reason to use an additional PDP-8 (not enough channels?).
    – dirkt
    Jun 8, 2019 at 17:29
  • 1
    16 buffers (=channels) plus no doubt some software-architectural limitations in the timesharing director. Terminal systems were all added by customers. Here is a contemporary paper about Eldon2 if you're interested. Jun 8, 2019 at 20:49
3

Keep in mind, you're talking about an age where no computer and even less it's devices where standardized components. Every manufacturer has his own ideas and parts where modified to fit.

And the Flexowriter did perform great here, as it was quite configurable, in some way even programmable by selecting/exchanging the parts. Hence the Flex part :))


What was the nature of the connection between the Flexowriter and the computer? Was this something we would recognize as a "serial interface" today?

Possible, depending on the computer, as there was also 6 bit parallel interface possible, but in most cases it would have been serial.

As well as the usual keyboard and printer functions, Flexowriters typically had integrated papertape readers and punches. Were these controlled by inband signaling (like XON/XOFF for ASCII Teletypes; not that I'm saying a Flex used ASCII) or "extra wires"?

Yes :))

As before. You have to look at the computer they where made for. Most would use some specific protocol to enable or disable the keyboard.

Flexowriters also seemed to have switches, some of which were clearly for local device control (start tape, perhaps) but others seem to be computer control (the LGP30 Flex had what looks like a "start computer" switch). How did they communicate with the computer?

Usually via separate wires.

If any answer is computer-specific, please indicate the computer you're talking about.

Lets skip that, as it would be an endless list with entries for next to all comuters using a Flexowriter.

2
  • 2
    I looked at a PDP-1 manual, and by the way it words the description of the appropriate IO instruction, it seems to hint at a 6-bit parallel interface being used on that machine. Admittedly this is programming-level doc, not hardware-level doc, though in this case the two things aren't entirely separated the way we expect them to be today. Jun 8, 2019 at 15:26
  • @another-dave true, but then again, that's only one machine of one manufacturer. The world was way more colourful back then :)
    – Raffzahn
    Sep 5, 2020 at 8:18
1

I'm adding my own answer as a sort of summary of what I got from the other two excellent answers.

  1. It depends on the computer, though the evidence favours a parallel interface for tightly-integrated consoles.

  2. Auxiliary device control, and computer control functions, were conveyed by specific signals rather than (say) in-band control characters.

I'd like to "accept" both the previous answers, but alas...

1
  • Taken together, this has been a great discussion showing that transitional period between everything-special-purpose and beginnings of standardization. Ken Shirriff has an excellent exposition of IBM 029 keypunch, electromechanical encoding with photos and diagrams: righto.com/2017/12/repairing-1960s-era-ibm-keypunch.html Jun 10, 2019 at 1:14
1

For the Librascope/Royal Precision/General Precision computers, i.e. the LGP-30 and its little and big brothers (LGP-21 and RPC-4000), the actual data transmission is 6 bit parallel. Two sets of 6 data lines, to and from the Flexowriter respectively.

This is complemented by a handful of handshake and sync signals, to indicate that characters are available, trigger their processing, or signal that an "end of input" code has been entered by the user or read by the paper tape reader. Some of the Flexowriter's handshake signals are directly derived from its mechanical camshaft. So they expose an amount of machine detail to the computer which is unexpected by today's standards.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .